matrix-spec-proposals/proposals/1772-groups-as-rooms.md

Proposal for groups as rooms (take 2)

This obsoletes MSC1215.

Problem

The current groups API has some serious issues:

• It is a large API surface to implement, maintain and spec - particularly for all the different clients out there.
• Much of the API overlaps significantly with mechanisms we already have for managing rooms:
  • Tracking membership identity
  • Tracking membership hierarchy
  • Inviting/kicking/banning user
  • Tracking key/value metadata
• There are membership management features which could benefit rooms which would also benefit groups and vice versa (e.g. "auditorium mode")
• The current implementations on Riot Web/iOS/Android all suffer bugs and issues which have been solved previously for rooms.
  • no local-echo of invites
  • failures to set group avatars
  • ability to specify multiple admins
• It doesn't support pushing updates to clients (particularly for flair membership): https://github.com/vector-im/riot-web/issues/5235
• It doesn't support third party invites.
• Groups could benefit from other features which already exist today for rooms
  • e.g. Room Directories
• Groups are centralised, rather than being replicated across all participating servers.

Solution

Represent groups by rooms rather than a custom first-class entity.

We reserve aliases which begin with a + to represent groups - e.g. the room for group +test:example.com is #+test:example.com.

We introduce m.room.group and m.room.subgroup events which define the rooms and subgroups within the group. A present: true key is included to distinguish from a deleted state event. Something like:

{
    "type": "m.room.group",
    "state_key": "#room1:example.com",
    "contents": {
        "present": true
    }
}

{
    "type": "m.room.group",
    "state_key": "#room2:example.com",
    "contents": {
        "present": true,
        "autojoin": true
    }
}

{
    "type": "m.room.subgroup",
    "state_key": "+something:example.com",
    "contents": {
        "present": true
    }
}
{
    "type": "m.room.subgroup",
    "state_key": "+otherthing:example.com",
    "contents": {
        "present": true
    }
}

Name, Topic, Membership etc share the same events as a normal room.

The flair image for a group is given by the room avatar. (In future it might preferable to use hand-crafted small resolution images: see matrix-doc#1778.

Long description requires a new event: m.room.description. This can also be used for rooms as well as groups.

Groups may be nested, and membership of groups is defined as the union of the members of the group room and its subgroups. If +top:example.com has two subgroups, the user membership of +top:example.com is the union of the subgroups and the group itself. This allows hierarchies of groups & users to be defined.

Clients peek in rooms (recursing into subgroups as needed) in order to determine group membership.

Invites, 3PID invites, Power Levels etc all work as for a normal room.

Normal messages within the room could be showed and used as a 'lobby' area for the given group.

This requires no server changes at all, other than better support for peeking (see Dependencies below), and could allow the existing /groups API to be deprecated and removed outright.

ACLs

Currently the group server has total control over specifying the list of users who may be present in a group, as seen by a given querying user. In other words, arbitrary users can see entirely different views of a group at the server's discretion.

Whilst this is very powerful for mapping arbitrary group structures into Matrix, it may be overengineered.

Instead, the common case is wanting to define a group where some users are publicly visible as members, and others are not. This is what the current use cases require today. A simple way of achieving would be to create a subgroup for the private members - e.g. have +sensitive:matrix.org and +sensitive-private:matrix.org. The membership of +sensitive-private:matrix.org is set up with m.room.join_rules to not to allow peeking; you have to be joined to see the members, and users who don't want to be seen by the public to be member of the group are added to the subgroup.

XXX: is there a use case today for having a group where users are unaware of the other users' membership? e.g. if I am a member of +scandalous:matrix.org should i have a way to stop other members knowing that I am? One solution here could be "auditorium mode", where users cannot see other users' identities (unless they speak). This could be added later, however, and would also be useful for normal rooms.

Flair

A proposal for how to safely determine user flair is:

• User publishes the groups they wish to announce on their profile (MSC1769 as a m.flair state event: it lists the groups which they are advertising.

• When a client wants to know the current flair for a set of users (i.e. those which it is currently displaying in the timeline), it peeks the profile rooms of those users. However, we can't trust the flair which the users advertise on the profile - it has to be cross-referenced from the memberships of the groups in question.

To do this cross-referencing, options are:

1. The client checks the group membership (very inefficient, given the server could/should do it for them), or...
2. The server checks the group membership by peeking the group and somehow decorates the m.flair event as validated before sending it to the client. This is also inefficient, as it forces the server to peek a potentially large group (unless we extend federation to allow peeking specific state events)
3. The origin m.flair event includes the event_id of the user's m.room.membership event in the group. The server performing the check can then query this specific event from one of the servers hosting the group-room, and we perhaps extend the S2S API to say whether a given state event is current considered current_state or not. If the m.room.membership event is confirmed as current, then the m.flair is decorated as being confirmed.

Of these, option 3 feels best?

Dependencies

This needs peeking to work effectively over the CS API (MSC1776).

This needs peeking to work effectively over federation (e.g. by having servers join remote rooms as @null:example.com in order to participate in them for peeking purposes - MSC1777).

These dependencies are shared with profiles-as-rooms (MSC1769).

Security considerations

XXX: how do we stop cycles & recursion abuse of the subgroups?

Tradeoffs

This consciously sacrifices the ability to delegate group lookups through to a centralised group server. However, group data can already be stale as we rely on cached attestations from federated servers to apply ACLs even if the remote server is not available. So this isn’t much worse than eventually consistent group membership as you’d find in a room.

It also means that large groups have to be bridged in their entirety into the room, rather than querying/searching incrementally. This is something we should fix for bridged rooms in general too, however.

This also consciously sacrifices the ability for a group server to provide different 'views' of groups to different querying users, as being overengineered. Instead, all common use cases should be modellable by modelling group membership as room membership (nesting if required).

Issues

How does this work with MSC1228 (removing MXIDs)?

Unstable prefix

While this proposal is not in a published version of the specification, implementations should use org.matrix.msc1772 to represent the m namespace. For example, m.room.subgroup becomes org.matrix.msc1772.room.subgroup.

History

• This replaces MSC1215: https://docs.google.com/document/d/1ZnAuA_zti-K2-RnheXII1F1-oyVziT4tJffdw1-SHrE
• Other thoughts that led into this are at: https://docs.google.com/document/d/1hljmD-ytdCRL37t-D_LvGDA3a0_2MwowSPIiZRxcabs